IPMiner: hidden ncRNA-protein interaction sequential pattern mining with stacked autoencoder for accurate computational prediction

Table 4 The predicted performance of trained model from RPI488 on NPInter2.0, RPI367 and RPIntDB dataset

Dataset	Organism	Total # of	Predicted # of
		ncRNA-protein	ncRNA-protein
NPInter2.0	Homo sapiens	6,975	6,809 (97.6 %)
	Caenorhabditis elegans	36	22 (61.1 %)
	Mus musculus	2,198	2,115 (96.2 %)
	Drosophila melanogaster	91	88 (96.7 %)
	Saccharomyces cerevisiae	910	860 (94.5 %)
	Escherichia coli	202	176 (87.1 %)
	Total	10,412	10,070 (96.7 %)
RPI367	Homo sapiens	148	132 (89.2 %)
	Caenorhabditis elegans	2	2 (100.0 %)
	Mus musculus	46	34 (73.9 %)
	Drosophila melanogaster	26	24 (92.3 %)
	Saccharomyces cerevisiae	119	117 (98.3 %)
	Escherichia coli	25	21 (84.0 %)
	Total	366	330 (90.1 %)
RPIntDB	Total	44,586	38,522 (86.4 %)

For NPInter2.0, RPI-Pred can predict 90 % of total interactions [13]. If proteins and RNAs in a pair are obsolete, then this pair will be removed. For example, in RPI367, protein O16646 is obsolete in UniProtKB, and ncRNA u1136 interacts with O16646, this pair was removed in RPI367. In RPIntDB, there is no organism information for some interaction pairs, so we only report the total prediction accuracy

ISSN: 1471-2164